Mining Scholarly Publications for Scientific Knowledge Graph Construction

Abstract

In this paper, we present a preliminary approach that uses a set of NLP and Deep Learning methods for extracting entities and relationships from research publications and then integrates them in a Knowledge Graph. More specifically, we i) tackle the challenge of knowledge extraction by employing several state-of-the-art Natural Language Processing and Text Mining tools, ii) describe an approach for integrating entities and relationships generated by these tools, and iii) analyse an automatically generated Knowledge Graph including 10, 425 entities and 25, 655 relationships in the field of Semantic Web.

Generating knowledge graphs by employing Natural Language Processing and Machine Learning techniques within the scholarly domain

“…When the value is closer to 1, it means that the clusters are well separated; when the value is closer to 0, it might be difficult to detect the decision boundary; when the value is closer to -1, it means that elements assigned to a cluster might have been erroneously assigned. Its formula is computed as shown in (2), where given a cluster c, w(c) represents the average dissimilarity of elements in c, and o(c) is the lowest average dissimilarity of elements of c to any other cluster.…”

“…Despite the large number and variety of tools and services available today for exploring scholarly data, current support is still very limited in the context of sensemaking tasks that require a comprehensive and accurate representation of the entities within a domain and their semantic relationships. This raises the need of more flexible and fine-grained scholarly data representations that can be used within technological infrastructures for the production of insights and knowledge out of the data [2,3,4]. Kitano [5] proposed a similar and more ambitious vision, suggesting the development of an artificial intelligence system able to make major scientific discoveries in biomedical sciences and win a Nobel Prize.…”

Multimodal Knowledge Graph for Deep Learning Papers and Code

“…After alignment we get a RDF knowledge graph which can be stored and queried in any triple store. All our code and documentation can be found online 3 . In the following sections we describe the different components of our system from the ontology design to the different types of knowledge graph extractors.…”

“…Since all of these elements are connected to a single paper through the Publication entity, we perform a local alignment by looking for entities that are named similarly across modalities and then linking them together. • Global graph alignment: We have decided to use an external well-curated knowledge graph called the Computer Science Ontology (CSO) -a large scale research ontology that categorizes 16 million publications, mainly in the field of computer science as a taxonomy [3]. Out of the 14K topics and 163K relationships, we mapped all our instances of concepts from the text extraction (i.e., the entities of the text extraction to their respective CSO components).…”

AI-KG: An Automatically Generated Knowledge Graph of Artificial Intelligence

“…An alternative vision, that is gaining traction in the last few years, is to generate a semantically rich and interlinked description of the content of research publications [13,29,7,24]. Integrating this data would ultimately allow us to produce large scale knowledge graphs describing the state of the art in a field and all the relevant entities, e.g., tasks, methods, metrics, materials, experiments, and so on.…”